Inclusion and Exclusion Criteria(Defining the Study Population)

• General considerations

• Run-in periods (enrichment)

• Regression toward the mean

Some Issues in Selecting Patients for Clinical Trials

• Broad or narrow criteria

• Recruitment and patient logs

• Nature of clinical sites (patient to research or research to patients)

Relationship of Study Sample to Study Population and Population at Large

Population at Large

Population withoutCondition

Population with Condition

With Conditionbut Ineligible

Study Population

Eligible butnot Enrolled

Study Sample

Source: Friedman, Furberg and DeMets.

Definition ofCondition

Entry Criteria

Enrollment

Internal and External Validity• Internal validity – can you attribute the

differences among treatment groups to the treatments studied– Randomization– Blinding– Sample size– Random and systematic errors minimized

• External validity – can you generalize the results to patients outside the trial (the same or larger target population); to whom to the results apply?– Inclusion criteria– Characteristics of those randomized– Setting of trial– Outcomes assessed and duration of follow-up

The Study Population is Defined byInclusion and Exclusion Criteria

Eligibility Criteria:– Should be defined in advance– No exceptions should be made

unless a priori stated–Characterize population (e.g.,

labeling indication for a new drug/device)• Impact of results (assessment by

others)• Replication of study

Considerations in DevelopingEligibility Criteria

1. Enroll patients who have potential to benefit from intervention

2. Enroll patients for whom there is a high probability of detecting results of intervention

3. Do not enroll patients to whom intervention is potentially harmful

4. Do not enroll patients at high risk of conditions which preclude ascertainment of endpoints

5. Do not enroll patients who cannot comply with protocol

Friedman, Furberg, DeMets: Fundamentals of Clinical Trials, Chapter 4

Selection of Patients for a Trial

• Drug clearly indicated based on current knowledge

Exclude

• Drug effect unknown:– But likely to be beneficial to certain patients

Include*

– And outcome uncertain in certain patients

? Include

– But theoretical possibility of harm or little hope

?? Include* of benefit

• Special exclusions (competing risk, poor

Exclude compliance, inability to ascertain endpoint)

• Drug contraindicated based on current knowledge

Exclude

Yusuf et al., Stat Med, 9:73-86, 1990.

* Include such patients with explicit prior statement of a subgroup hypothesis

Advantages and Disadvantagesof Opposing Selection Strategies

Highly restrictive selection criteria• Advantages

– Provides more precise comparison of the test and control treatments– Results of the trial less likely to be affected by population variability

• Disadvantages– Increases the cost and time required for patient recruitment– Limits the generalizability of the study findings

Minimally restrictive selection criteria• Advantages

– Makes patient recruitment easier– Provides base for wider generalization of findings

• Disadvantages– May obscure treatment effects because of variability in composition of study

population– Results of the trial may be confusing, especially if an observed effect

appears to be associated with a subgroup of patients in the study and the subgroup is too small to yield a reliable treatment comparison

Meinert C., Clinical Trials. Design, Conduct and Analysis.

Usually Favor Minimal Restrictions for Large Outcome Trials

• A priori, often not clear who will do best

• Rare that there are large subgroup differences of a qualitative nature

• Faster answer to question

• With few restrictions, risk stratification after the trial is completed can be performed (more information to formulate treatment guidelines; cannot assess consistency of results in a subgroup if they were excluded)

Does it Work?“Practical Clinical Trials: Increasing the Value

of Clinical Research for Decision Making in Clinical and Health Policy”

• Treatments that mimic use in practice

• Diverse patient population; minimal exclusions

• Many clinical sites; heterogeneous practice settings

• Multiple, easily ascertained, clinically relevant endpoints

Tunis SR et al, JAMA 2003. Center for Medicare and Medicaid Services and

Agency for Healthcare Research and Quality

Explanatory and Pragmatic Trials

• Explanatory (“laboratory” conditions)– Strict selection criteria aimed at creating a

homogenous patient population

• Pragmatic (“normal” conditions)– Heterogeneous patient population, one where

might expect more non-adherence

Schwartz & Lellouch J Chronic Dis 1967)

Recruitment Considerations• Always more difficult than anticipated (need a

backup plan to the backup plan)

• Easier with broad eligibility criteria (e.g., large simple trials)

• Yield not 100%– Eligibility criteria (age, prior history, prior treatment, etc.)– Exclusion criteria– Physician refusal– Patient refusal

• MRFIT screened 361,662 men to identify 12,866

• Must closely monitor to ensure target is met

Patient LogsShould You Count Those Not Included?

Possible Advantages• May provide insights on how to improve slow enrollment• May allow more reasonable recommendations concerning

findings – generalizability• Allows an assessment of how many eligible patients are

being enrolledDisadvantages• Time-consuming• Often uncertain where the counting starts• Issue of generalizability usually cannot be addressed with

logs alone

Randomized Trialof Extracranial-Intracranial Arterial Bypass

Patients Undergoing Randomization versus Patients Operated on Outside of the Trial in 57 of 71 Participating Centers

United States 30/33 252 237 1351Europe 15/19 203 190 741Japan 12/14 105 90 480Canada 5* 94 84 ?

TOTAL 62 654 601 2572

Location

No. ofCenters

Surveyed/Total

Randomizedto MedicalTreatment

Randomizedto Surgery

Operatedon Outside

the Trialno. of patients

* These centers were not surveyed; randomization was assumed to have been complete.

Sundt, TM. N Eng J Med 316:814-16, 1987.

MRFIT Recruitment• 22 centers (specially funded to carry

out the study)• Sources of 12,866 participants (no.

centers using source)– Census tract listings (door-to-door) (3)– Industry/government (15)– HMOs (2)– Union lists (10)– Churches (7)– Shopping centers (10)– Civic clubs (8)– Other (14)

ESPRITHIV – IL-2

• 243 clinical sites

• 25 countries

• 4,150 patients

ELITE IICHF – Losartan

• 288 sites

• 45 countries

• 3,152 patients

Some trials require many sites and different types of sites

Advantages of Multi-Center Trials

• SAMPLE SIZE!

• Faster enrollment and therefore faster answer to the research question

• External validity (generalizability)– diversity of patients– convenience to patients– diversity of clinical sites

• Credibility (each site is a replicate)

• More clinicians involved in research

Run-in Periods

Def. A period before randomization in which patients are given a treatment (active or

placebo) to establish eligibility.

Run-ins involve compromises between internal and external validity and are

often considered for reasons of efficiency.

Pablo-Mendez A, JAMA 1998

Rationale for Run-In Periods

1. Expected compliance is low;2. Substantial behavior change is required (SNaP);3. Contact with participants is infrequent (Physician’s Health

Study);4. Possibility of side effects to experimental treatment;5. A large number of “placebo responders” are expected

(e.g., pain study);6. It is necessary to assure disease status is stable (or

washout);7. Efficacy of treatment on intermediate response variables

is required, thereby enhancing clinical applicability (CAST).

May be active or placebo; generally considered when:

ITT and Run-Ins

If the primary interest (primary estimand) is the difference in outcome between treatment and

control groups among patients who can tolerate the treatment, it is better to use a run-in period and ITT analysis than to try and estimate this

estimand using only the patients who tolerated the treatment after randomization.

Run-In Periods in Trials - Examples

• To enhance clinical applicability (CAST)

• To screen for placebo response (pindolol + fluoxetine versus placebo + fluoxetine for depression)

• To assess tolerability of active treatment (carvedilol for the treatment of heart failure)

• To screen for non-adherence (Physician’s Health Study used carotene placebo and open-label aspirin during run-in)

All have the potential to increase power or reduce sample size requirements

Potential Problems with Run-In Periods

• Carryover effects

• Blinding jeopardized

• Interpretation of results have to be qualified

• Strategy for enrichment wrong

Leber PD and Davis CS, Cont Clin Trials 1998

Enrichment Designs

• Enrichment designs refers to studies that aim to increase power by choosing patients that are more likely to show benefit than a broader population (personalized medicine).– In heart failure and hypertension – initial studies

were in more advance patients– Active run-in periods (e.g. CAST) or randomized

withdrawal studies– Biomarkers and genetic traits that predict

response (e.g., expression of HER-2 in tumors in women with metastatic breast cancer)

Temple R, Commun Statist Theory Meth, 1994

An Example of Enrichment That Backfired

• Tacrine for the treatment of Alzheimer’s disease (N Engl J Med 1992;327:1253-1259)– Patients first randomized to 2 doses of tacrine (10 or 20 mg

four times a day) or placebo (titration or enrichment phase) for 6 weeks (crossover).

– Following 2 week placebo washout period, responders in first phase randomized to their best dose or placebo for 6 weeks (parallel group).

– This was followed by an active treatment phase (uncontrolled).

• Concerns: Carryover effects, withdrawal effects, maintenance of blind, and generalizability

See also Leber and Davis, Cont Clin Trials, 1998 and Encyclopedia of Biopharmaceutical Statistics

Factors Influencing Efficiency of Targeted Therapy (Enrichment) Design• Heterogeneity of effect across subgroups

(e.g., test positive patients expected to do much better than test-negative patients)

• Prevalence of factor describing responder group

• Sensitivity and specificity of test to define the responders

There is often uncertainty. Thus, may first have to determine efficacy of treatment in both test-positive and test negative patients.

Maitournam and Simon, Stat Med 2005

Summary

• In most trials eligibility requirements are too restrictive (an opinion)

• Arguments concerning logs versus no logs arise because of restrictions

• Trial entry procedures are simplified with broad inclusion criteria

• Study design teams should carefully consider reasons for exclusion

• Exemptions should not be granted

• We need to determine way of efficiently involving more clinicians in the community in research

• There are advantages and disadvantages to run-ins and enrichment designs. The latter are being actively pursued as part of a “personalized medicine” research agenda.